For separating the patches, why not just make the Scheme API code pass a NULL GError in the main patch, since that's a simple and obviously correct one-line change that keeps the current behaviour? In a subsequent patch, you could then pass a non-NULL pointer, check the error condition, and do the other API updates.
The branch sounds like an excellent idea.
For separating the patches, why not just make the Scheme API code pass a NULL GError in the main patch, since that's a simple and obviously correct one-line change that keeps the current behaviour? In a subsequent patch, you could then pass a non-NULL pointer, check the error condition, and do the other API updates.